Direct answer
ASO keyword research is the process of turning what users actually search for into a ranked list of phrases the listing can realistically own, then assigning each phrase to a specific field: title, subtitle, keyword field, or description proof points. A workable workflow starts from seed terms (what the app does, what problem it solves, what competitors are called), expands them through store autocomplete and competitor listings, then filters by relevance before volume β a high-volume keyword the app cannot honestly serve produces installs that churn and reviews that hurt. The output of keyword research is not a spreadsheet of 500 terms; it is a small set of keyword clusters with one decided owner field each, ready to drive metadata and screenshot copy. Teams fail here when research happens once at launch and never again, or when the keyword list never actually changes what gets written.
Where keywords come from
| Source | What it gives you | Watch out for |
|---|---|---|
| Store autocomplete | Real query phrasing users type | Reflects popularity, not relevance to your app |
| Competitor titles and subtitles | Terms the category already competes on | Copying positioning instead of finding gaps |
| Reviews (yours and competitors') | The vocabulary users naturally use | Small samples can overweight edge cases |
| Support tickets and onboarding questions | Problem language before users learn your terms | Needs translation into search phrasing |
Recommended flow
1. Collect seed terms from the product, not the brand deck
List what the app does in plain words, the problems it solves, and the categories users would place it in. Brand language comes later; research starts from user language.
2. Expand each seed through autocomplete and competitors
Type each seed into store search and record the suggestions. Read the top ten competitors' titles and subtitles for the same seeds. This surfaces both phrasing variants and adjacent intents.
3. Filter by relevance before volume
Cut every term the app cannot honestly serve. Ranking for a query you disappoint converts badly and invites negative reviews that suppress every other keyword.
4. Cluster the survivors by intent
Group terms that express the same intent ("habit tracker", "habit app", "daily habits") into clusters. Each cluster gets one primary phrase and several variants.
5. Assign each cluster to a field
The strongest cluster goes to the title, the second to the subtitle, variants to the keyword field, and supporting intents become description and screenshot proof points. A cluster without an assigned field is research that changed nothing.
6. Revisit after every release cycle
Rankings, competitors, and autocomplete all move. Re-run the expansion step at least once per release cycle and record what changed, so metadata edits are grounded in movement rather than mood.
Common failure modes
Research ends as a spreadsheet, not a decision
A 500-row keyword sheet that never dictates what the title says is decoration. The deliverable is clusters with owner fields, not row count.
Volume outranks relevance
Chasing the biggest keyword in the category puts a niche app in competition it cannot win, on a promise it cannot keep.
Bilingual research is one language translated
Chinese users do not search translated English phrases. Each locale needs its own autocomplete expansion and its own clusters, aligned in strategy but researched independently.
Keyword research checklist
- Seed terms describe the product in user language, not brand language.
- Every seed was expanded through autocomplete and competitor listings.
- Irrelevant high-volume terms were cut, with the reason recorded.
- Surviving terms are clustered by intent with one primary phrase each.
- Every cluster has an assigned field: title, subtitle, keyword field, or description.
- Each locale was researched in its own language.
Operating rule
If a keyword cluster cannot name the field it owns and the claim it supports, it has not finished research β it is still a list.
Why this matters in App Store Helper
App Store Helper keeps keyword decisions attached to the project instead of a detached spreadsheet: clusters feed title and subtitle drafts, variants flow into the keyword field, and the same hierarchy drives screenshot headlines. When research is revisited after a release, the diff between old and new metadata stays reviewable in one place.